Prediction Accuracy Of Classifier Research Articles

Cancer Diagnosis by Gene-Environment Interactions via Combination of SMOTE-Tomek and Overlapped Group Screening Approaches with Application to Imbalanced TCGA Clinical and Genomic Data

The complexity of cancer development involves intricate interactions among multiple biomarkers, such as gene-environment interactions. Utilizing microarray gene expression profile data for cancer classification is anticipated to be effective, thus drawing considerable interest in the fields of bioinformatics and computational biology. Due to the characteristics of genomic data, problems of high-dimensional interactions and noise interference do exist during the analysis process. When building cancer diagnosis models, we often face the dilemma of model adaptation errors due to an imbalance of data types. To mitigate the issues, we apply the SMOTE-Tomek procedure to rectify the imbalance problem. Following this, we utilize the overlapping group screening method alongside a binary logistic regression model to integrate gene pathway information, facilitating the identification of significant biomarkers associated with clinically imbalanced cancer or normal outcomes. Simulation studies across different imbalanced rates and gene structures validate our proposed method’s effectiveness, surpassing common machine learning techniques in terms of classification prediction accuracy. We also demonstrate that prediction performance improves with SMOTE-Tomek treatment compared to no imbalance treatment and SMOTE treatment across various imbalance rates. In the real-world application, we integrate clinical and gene expression data with prior pathway information. We employ SMOTE-Tomek and our proposed methods to identify critical biomarkers and gene-environment interactions linked to the imbalanced binary outcomes (cancer or normal) in patients from the Cancer Genome Atlas datasets of lung adenocarcinoma and breast invasive carcinoma. Our proposed method consistently achieves satisfactory classification accuracy. Additionally, we have identified biomarkers indicative of gene-environment interactions relevant to cancer and have provided corresponding estimates of odds ratios. Moreover, in high-dimensional imbalanced data, for achieving good prediction results, we recommend considering the order of balancing processing and feature screening.

A Performance and Data-Driven Method for Optimization of Traditional Courtyards

As urbanization and rapid industrialization accelerate, rural areas face increasing pressure on resources and the environment, leading to challenges such as energy waste and reduced comfort. Traditional village planning and design methods are based on economic benefits and often lack consideration of climate adaptability. To address these issues, a comprehensive assessment of building and courtyard performance should be introduced early in the planning of traditional villages. This approach can better adapt the buildings to their climatic conditions. Introducing relevant performance indicators, such as outdoor comfort, indoor lighting, and building energy consumption, at the initial design stage is crucial. This article employs performance-based multi-objective optimization algorithms and machine learning techniques to investigate the design workflow of courtyards and their combinations. The goal is to enhance planners’ design efficiency in village planning by integrating data-driven and performance-driven methods. The research results show that during the performance-driven phase, by adjusting the spatial morphology and architectural parameters, the performance of the courtyard significantly improved compared to the baseline model. Energy efficiency increased by 32.3%, the physiological equivalent temperature (PET) comfort time ratio in winter was enhanced by 8.3%, and the ratio in summer increased by 3.8%. During the data-driven phase, the classification prediction accuracy of courtyard performance can reach 83%, and the F1 score is 0.81. In the project validation phase, it has also been proven that the performance of different plans can be quickly verified. Compared to the base’s original status, the design solutions’ performance score can be improved from 59.12 to 85.62. In summary, this workflow improves the efficiency of the interaction between design decisions and performance evaluation in the conceptual stage of village planning, providing a solid foundation for promoting subsequent solutions.

A Performance and Data-Driven Method for Optimization of Traditional Courtyards

As urbanization and rapid industrialization accelerate, rural areas face increasing pressure on resources and the environment, leading to challenges such as energy waste and reduced comfort. Traditional village planning and design methods are based on economic benefits and often lack consideration of climate adaptability. To address these issues, a comprehensive assessment of building and courtyard performance should be introduced early in the planning of traditional villages. This approach can better adapt the buildings to their climatic conditions. Introducing relevant performance indicators, such as outdoor comfort, indoor lighting, and building energy consumption, at the initial design stage is crucial. This article employs performance-based multi-objective optimization algorithms and machine learning techniques to investigate the design workflow of courtyards and their combinations. The goal is to enhance planners’ design efficiency in village planning by integrating data-driven and performance-driven methods. The research results show that during the performance-driven phase, by adjusting the spatial morphology and architectural parameters, the performance of the courtyard significantly improved compared to the baseline model. Energy efficiency increased by 32.3%, the physiological equivalent temperature (PET) comfort time ratio in winter was enhanced by 8.3%, and the ratio in summer increased by 3.8%. During the data-driven phase, the classification prediction accuracy of courtyard performance can reach 83%, and the F1 score is 0.81. In the project validation phase, it has also been proven that the performance of different plans can be quickly verified. Compared to the base’s original status, the design solutions’ performance score can be improved from 59.12 to 85.62. In summary, this workflow improves the efficiency of the interaction between design decisions and performance evaluation in the conceptual stage of village planning, providing a solid foundation for promoting subsequent solutions.

Diffusion model enables quantitative CBF analysis of Alzheimer's Disease.

Cerebral blood flow (CBF) measured by arterial spin labeling (ASL) is a promising biomarker for Alzheimer's Disease (AD). ASL data from multiple vendors were included in the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. However, the M0 images were missing in Siemens ASL data, prohibiting CBF quantification. Here, we utilized a generative diffusion model to impute the missing M0 and validated generated CBF data with acquired data from GE. A conditional latent diffusion model was trained to generate the M0 image and validate it on an in-house dataset ( N =55) based on image similarity metrics, accuracy of CBF quantification, and consistency with the physical model. This model was then applied to the ADNI dataset (Siemens: N =211) to impute the missing M0 for CBF calculation. We further compared the imputed data (Siemens) and acquired data (GE) regarding regional CBF differences by AD stages, their classification accuracy for AD prediction, and CBF trajectory slopes estimated by a mixed effect model. The trained diffusion model generated the M0 image with high fidelity (Structural similarity index, SSIM=0.924±0.019; peak signal-to-noise ratio, PSNR=33.348±1.831) and caused minimal bias in CBF values (mean difference in whole brain is 1.07±2.12ml/100g/min). Both generated and acquired CBF data showed similar differentiation patterns by AD stages, similar classification performance, and decreasing slopes with AD progression in specific AD-related regions. Generated CBF data also improved accuracy in classifying AD stages compared to qualitative perfusion data. This study shows the potential of diffusion models for imputing missing modalities for large-scale studies of CBF variation with AD.

Machine Learning Model Stability for Sub-Regional Classification of Barossa Valley Shiraz Wine Using A-TEEM Spectroscopy.

With a view to maintaining the reputation of wine-producing regions among consumers, minimising economic losses caused by wine fraud, and achieving the purpose of data-driven terroir classification, the use of an absorbance-transmission and fluorescence excitation-emission matrix (A-TEEM) technique has shown great potential based on the molecular fingerprinting of a sample. The effects of changes in wine composition due to ageing and the stability of A-TEEM models over time had not been addressed, however, and the classification of wine blends required investigation. Thus, A-TEEM data were combined with an extreme gradient boosting discriminant analysis (XGBDA) algorithm to build classification models based on a range of Shiraz research wines (n = 217) from five Barossa Valley sub-regions over four vintages that had aged in bottle for several years. This spectral fingerprinting and machine learning approach revealed a 100% class prediction accuracy based on cross-validation (CV) model results for vintage year and 98.8% for unknown sample prediction accuracy when splitting the wine samples into training and test sets to obtain the classification models. The modelling and prediction of sub-regional production area showed a class CV prediction accuracy of 99.5% and an unknown sample prediction accuracy of 93.8% when modelling with the split dataset. Inputting a sub-set of the current A-TEEM data into the models generated previously for these Barossa sub-region wines yielded a 100% accurate prediction of vintage year for 2018-2020 wines, 92% accuracy for sub-region for 2018 wines, and 91% accuracy for sub-region using 2021 wine spectral data that were not included in the original modelling. Satisfactory results were also obtained from the modelling and prediction of blended samples for the vintages and sub-regions, which is of significance when considering the practice of wine blending.

An early force prediction control scheme using multimodal sensing of electromyography and digit force signals

Different grasping gestures result in the change of muscular activity of the forearm muscles. Similarly, the muscular activity changes with a change in grip force while grasping the object. This change in muscular activity, measured by a technique called Electromyography (EMG) is used in the upper limb bionic devices to select the grasping gesture. Previous research studies have shown gesture classification using pattern recognition control schemes. However, the use of EMG signals for force manipulation is less focused, especially during precision grasping.In this study, an early predictive control scheme is designed for the efficient determination of grip force using EMG signals from forearm muscles and digit force signals. The optimal pattern recognition (PR) control schemes are investigated using three different inputs of two signals: EMG signals, digit force signals and a combination of EMG and digit force signals. The features extracted from EMG signals included Slope Sign Change, Willison Amplitude, Auto Regressive Coefficient and Waveform Length. The classifiers used to predict force levels are Random Forest, Gradient Boosting, Linear Discriminant Analysis, Support Vector Machines, k-nearest Neighbors and Decision Tree. The two-fold objectives of early prediction and high classification accuracy of grip force level were obtained using EMG signals and digit force signals as inputs and Random Forest as a classifier. The earliest prediction was possible at 1000 ms from the onset of the gripping of the object with a mean classification accuracy of 90 % for different grasping gestures.Using this approach to study, an early prediction will result in the determination of force level before the object is lifted from the surface. This approach will also result in better biomimetic regulation of the grip force during precision grasp, especially for a population facing vision deficiency.

A colonial serrated polyp classification model using white-light ordinary endoscopy images with an artificial intelligence model and TensorFlow chart

In this study, we implemented a combination of data augmentation and artificial intelligence (AI) model—Convolutional Neural Network (CNN)—to help physicians classify colonic polyps into traditional adenoma (TA), sessile serrated adenoma (SSA), and hyperplastic polyp (HP). We collected ordinary endoscopy images under both white and NBI lights. Under white light, we collected 257 images of HP, 423 images of SSA, and 60 images of TA. Under NBI light, were collected 238 images of HP, 284 images of SSA, and 71 images of TA. We implemented the CNN-based artificial intelligence model, Inception V4, to build a classification model for the types of colon polyps. Our final AI classification model with data augmentation process is constructed only with white light images. Our classification prediction accuracy of colon polyp type is 94%, and the discriminability of the model (area under the curve) was 98%. Thus, we can conclude that our model can help physicians distinguish between TA, SSA, and HPs and correctly identify precancerous lesions such as TA and SSA.

Efficient Ensemble Broad Learning System Based on Dropout and DropConnect

Broad Learning System is an emerged efficient algorithm for training single hidden layer feedforward neural network (SLFN) with fast speed and good generalization ability. However, it is very difficult to determine the appropriate broad learning system structure and broad learning system may perform overfitting due to the dependence between nodes in processing fully connected network. In order to deal with these problems, efficient ensemble broad learning system based on Dropout and DropConnect are proposed in this paper. The proposed Dropout Ensemble Broad Learning System randomly discards hidden nodes to improve diversity between individuals and reduce the synergy between nodes to improve prediction stability. The DropConnect Ensemble Broad Learning System randomly drops connect weights to generate more complementary models by adding input attribute disturbance. The experimental results and statistical analysis on UCI data sets confirm that the proposed method can solve the problem of model overfitting caused by the strong dependence between the nodes of ensemble broad learning system and also show that the proposed approaches outperform the original BLS on prediction stability and classification accuracy.

Structure Optimization and Data Processing Method of Electronic Nose Bionic Chamber for Detecting Ammonia Emissions from Livestock Excrement Fermentation.

In areas where livestock are bred, there is a demand for accurate, real-time, and stable monitoring of ammonia concentration in the breeding environment. However, existing electronic nose systems have slow response times and limited detection accuracy. In this study, we introduce a novel solution: the bionic chamber construction of the electronic nose is optimized, and the sensor response data in the chamber are analyzed using an intelligent algorithm. We analyze the structure of the biomimetic chamber and the surface airflow of the sensor array to determine the sensing units of the system. The system employs an electronic nose to detect ammonia and ethanol gases in a circulating airflow within a closed box. The captured signals are processed, followed by the application of classification and regression models for data prediction. Our results suggest that the system, leveraging the biomimetic chamber, offers rapid gas detection response times. A high classification prediction accuracy, with a determination coefficient R2 value of 0.99 for single-output regression and over 0.98 for multi-output regression predictions, is achieved by incorporating a backpropagation (BP) neural network algorithm. These outcomes demonstrate the effectiveness of the electronic nose, based on an optimized bionic chamber combined with a BP neural network algorithm, in accurately detecting ammonia emitted during livestock excreta fermentation, satisfying the ammonia detection requirements of breeding farms.

Leveraging Unidentified Metabolic Features for Key Pathway Discovery: Chemical Classification-driven Network Analysis in Untargeted Metabolomics.

Untargeted metabolomics using liquid chromatography-electrospray ionization-high-resolution tandem mass spectrometry (UPLC-ESI-MS/MS) provides comprehensive insights into the dynamic changes of metabolites in biological systems. However, numerous unidentified metabolic features limit its utilization. In this study, a novel approach, the Chemical Classification-driven Molecular Network (CCMN), was proposed to unveil key metabolic pathways by leveraging hidden information within unidentified metabolic features. The method was demonstrated by using the herbivore-induced metabolic response in corn silk as a case study. Untargeted metabolomics analysis using UPLC-MS/MS was performed on wild corn silk and two genetically modified lines (pre- and postinsect treatment). Global annotation initially identified 256 (ESI-) and 327 (ESI+) metabolites. MS/MS-based classifications predicted 1939 (ESI-) and 1985 (ESI+) metabolic features into the chemical classes. CCMNs were then constructed using metabolic features shared classes, which facilitated the structure- or class annotation for completely unknown metabolic features. Next, 844/713 significantly decreased and 1593/1378 increased metabolites in ESI-/ESI+ modes were defined in response to insect herbivory, respectively. Method validation on a spiked maize sample demonstrated an overall class prediction accuracy rate of 95.7%. Potential key pathways were prescreened by a hypergeometric test using both structure- and class-annotated differential metabolites. Subsequently, CCMN was used to deeply amend and uncover the pathway metabolites deeply. Finally, 8 key pathways were defined, including phenylpropanoid (C6-C3), flavonoid, octadecanoid, diterpenoid, lignan, steroid, amino acid/small peptide, and monoterpenoid. This study highlights the effectiveness of leveraging unidentified metabolic features. CCMN-based key pathway analysis reduced the bias in conventional pathway enrichment analysis. It provides valuable insights into complex biological processes.

Prediction of surface roughness using deep learning and data augmentation

PurposeSurface roughness has a serious impact on the fatigue strength, wear resistance and life of mechanical products. Realizing the evolution of surface quality through theoretical modeling takes a lot of effort. To predict the surface roughness of milling processing, this paper aims to construct a neural network based on deep learning and data augmentation.Design/methodology/approachThis study proposes a method consisting of three steps. Firstly, the machine tool multisource data acquisition platform is established, which combines sensor monitoring with machine tool communication to collect processing signals. Secondly, the feature parameters are extracted to reduce the interference and improve the model generalization ability. Thirdly, for different expectations, the parameters of the deep belief network (DBN) model are optimized by the tent-SSA algorithm to achieve more accurate roughness classification and regression prediction.FindingsThe adaptive synthetic sampling (ADASYN) algorithm can improve the classification prediction accuracy of DBN from 80.67% to 94.23%. After the DBN parameters were optimized by Tent-SSA, the roughness prediction accuracy was significantly improved. For the classification model, the prediction accuracy is improved by 5.77% based on ADASYN optimization. For regression models, different objective functions can be set according to production requirements, such as root-mean-square error (RMSE) or MaxAE, and the error is reduced by more than 40% compared to the original model.Originality/valueA roughness prediction model based on multiple monitoring signals is proposed, which reduces the dependence on the acquisition of environmental variables and enhances the model's applicability. Furthermore, with the ADASYN algorithm, the Tent-SSA intelligent optimization algorithm is introduced to optimize the hyperparameters of the DBN model and improve the optimization performance.

Land-use classification of Malaysian soils by ultra-high performance liquid chromatography (UHPLC)-based untargeted data combined with chemometrics for forensic provenance

Soil has become the most precious trace evidence attributed to its transference and persistence, but little is known about the capability of non-volatile organic fractions of soil in predicting land-use class. Hence, this study devoted to discriminating soils into four land-use classes based on chromatographic data and predictive modelling. Nine sites representing four different land-use classes were selected for collecting 73 soil samples via a grid method. Then, all the soil samples were dried, homogenized, sieved and eventually extracted using acetonitrile. The extracts were further analyzed via an isocratic elution in an ultra-high performance liquid chromatography (UHPLC) system. The pixel-level chromatograms were carefully assessed via diverse data preprocessing (DP) methods including baseline correction, normalization and derivative algorithms. Furthermore, the retention time (RT) window was also segmented into several sub-windows and thoroughly evaluated. Prediction accuracy of classification and regression tree (CART) modelling as estimated using 100 subsets of training samples on the corresponding testing and blind samples were ranked using TOPSIS method for identifying the most desired sub-window and DP strategy. The sub-RT window covering 0–5 min preprocessed via modified polynomial fitting algorithm followed by vector normalization emerged to be the most desired sub-dataset. The best sub-dataset modelled using CART and partial least squares-discriminant analysis (PLS2-DA) algorithms achieved 79.5 % and 97.4 % accuracy in predicting the land-use classes of 13 mixed samples. In conclusion, UHPLC-based fingerprint technique coupled with predictive modelling shows great potential in inferring land-use class of trace amount of soil.

Developing deep LSTMs with later temporal attention for predicting COVID-19 severity, clinical outcome, and antibody level by screening serological indicators over time.

The clinical course of COVID-19, as well as the immunological reaction, is notable for its extreme variability. Identifying the main associated factors might help understand the disease progression and physiological status of COVID-19 patients. The dynamic changes of the antibody against Spike protein are crucial for understanding the immune response. This work explores a temporal attention (TA) mechanism of deep learning to predict COVID-19 disease severity, clinical outcomes, and Spike antibody levels by screening serological indicators over time. We use feature selection techniques to filter feature subsets that are highly correlated with the target. The specific deep Long Short-Term Memory (LSTM) models are employed to capture the dynamic changes of disease severity, clinical outcome, and Spike antibody level. We also propose deep LSTMs with a TA mechanism to emphasize the later blood test records because later records often attract more attention from doctors. Risk factors highly correlated with COVID-19 are revealed. LSTM achieves the highest classification accuracy for disease severity prediction. Temporal Attention Long Short-Term Memory (TA-LSTM) achieves the best performance for clinical outcome prediction. For Spike antibody level prediction, LSTM achieves the best permanence. The experimental results demonstrate the effectiveness of the proposed models. The proposed models can provide a computer-aided medical diagnostics system by simply using time series of serological indicators.

Deep learning based prediction of traffic peaks in mobile networks

In mobile networks, it is essential to configure networks more efficiently to provide mobile users with services having better quality. For the adjacent cells, sometimes the mobile traffic concentrates in a single cell, leading to traffic imbalance; in many cases, the detection of imbalance is strongly related to the prediction of traffic peaks, which is extremely difficult because many peaks appear suddenly for no apparent reason. To better predict the peaks and traffic imbalance, we propose two novel mobile traffic predictors. The first is a Mixture of Experts (MoE) model which yields significantly better peak prediction along with excellent interpretability by establishing a cooperation mechanism between different experts, whereas the second predictor is a lightweight Multilayer Perceptron (MLP) which can obtain similar peak forecasting performance but operating more flexibly and consuming less computational power. The obtained predictions are then used to aid the predictive detection of traffic imbalance. To this end, we first perform a large-scale analysis of mobile traffic and then propose different approaches to detect the future traffic imbalance based on the predictions. To evaluate the performance of the proposed approaches, we have conducted extensive experiments on real-world mobile network datasets; all models are implemented in Python 3.8.8 with Pytorch 1.7.0, and they were run on a computer with i7-7700K CPU @ 4.2 GHz running Linux, 64 GB memory and a single Nvidia TITAN X Pascal 12GB GPU. The proposed models are compared with 10 widely used deep learning based predictors, and the results show that our approaches have significantly improved the sensitivity (peak prediction accuracy) from 39.3% to 58.5% (i.e., roughly a 50% increase) along with excellent interpretability. Moreover, the predictions are further used by the predictive detection approach, where the best multi-model approach improves the predictive classification accuracy of congestion (imbalance prediction accuracy) by 10.3% in contrast with the naive approach; this result also verifies the effectiveness of our peak predictions.

Research on the Reform and Practice of Informatization Mode of Adult Higher and Continuing Education Academic Records Management in the Context of Three-Whole Parenting

Abstract Information technology reform is an important way for schools to realize efficient management and information technology should be introduced into the management of student registration files to accelerate the process of informationization of registration file management. This paper builds an informatization database of adult higher continuing education student registration file management based on informatization technology and carries out the design of the main business process of student registration files. The multi-layer stacked LSTM network in the ELMo model is utilized to learn the preprocessing of the text of the school records data, and the CART decision tree algorithm is improved based on the reinforcement learning mechanism in order to improve the classification prediction accuracy of the decision tree in the dataset. In order to verify the feasibility of the school registration file informationization database constructed in this paper, a test analysis is conducted for the application of data mining technology and informationization database. The results show that the check accuracy of the CART decision tree algorithm improved by the reinforcement learning mechanism is around 80%, and the throughput of this paper’s database reaches 4827ops/s when the data writing volume is 1*105GB. When the number of threads reaches 128, the throughput of this paper’s informative database is around 5486ops/s, and when the data volume reaches 12*105 entries, the database in this paper the query response time is around 602ms. Under the background of three-round education, it is necessary to fully integrate information technology with the management of school registration files, which will lead to the innovation of the informationization mode of the management of school registration files in adult higher education.

Soong Classification Using Radiographs Only Moderately Correlates With Distal Radius Plate Position on Computed Tomography

Background: The Soong classification grades the prominence of volar locking plates used to treat distal radius fractures in relation to the volar rim. The basis of the classification scheme is that increasing plate prominence over the volar rim and distal radius watershed zone is associated with increased likelihood of flexor tendon irritation and need for plate removal. However, recent studies report mixed results on the predictive value of the Soong classification for these outcomes. We hypothesized that the decreased predictive accuracy of Soong classification is due to misclassification between Soong grades secondary to a suboptimal correlation between the Soong classification on radiographs (XRs) and computed tomography (CT). Methods: Fifty volarly displaced distal radius fractures treated with a volar locking plate in the international and publicly available ICUC database were reviewed. All cases with a postoperative XR and CT were included. Soong classification of the volar locking plate in relation to the volar ulnar rim was determined on both XR and CT by 2 independent, fellowship-trained hand surgeons using CT imaging as the gold standard. The distribution of Soong grades on XR and CT was compared using Pearson’s χ2 test, and correlation was calculated using the Matthews correlation coefficient (MCC). A multi-class confusion matrix was used to calculate each grade’s positive predictive value (PPV). Results: We found an MCC of 0.65, indicating only moderate correlation between the 2 modalities. Per individual Soong grade, the PPV was the highest for grade 2 (0.96), with lower PPVs for grade 0 (0.63) and grade 1 (0.60). Conclusions: The distribution of Soong grades was significantly different when using XR versus CT ( P < .001). Level of Evidence: Oxford Centre for Evidence-Based Medicine, diagnostic, level 2b

Dynamic decision-making for inspecting the quality of treated sewage

Given the scarcity of fresh water, aquatic ecosystems influence climate change. Municipal wastewater pumped into the sewage treatment system could then begin to be re-utilized, balancing how fresh water can be reused with how to fully utilize the potential value of water resources. The crux of the drainage standard is that the decision-making model is accurate and effective for detecting the quality of treated sewage and improving sewage treatment measures. The proposed method called PRBF-SVM(Parallel Radial Basis Function Support Vector Machine) predicts the classification of disposed sewage and provides an effective and efficient solution for detecting water quality. To solve the decision-making problem in municipal wastewater systems, a relevant decision model is set up to combine multi-attributes, and parallel grid search and cross-validation are employed to obtain the optimal model on the training data. Experimental results show that the proposed method improves the accuracy of classification prediction, which is higher than that of the original algorithm. In addition, the performance of the run-time and accuracy are perfect, compared to logistic regression, decision tree, random forest, KNeighbours, XGboost, and AdaBoost methods on the same dataset. PRBF-SVM can be used to determine whether sewage is disposed of to meet the post-treatment drainage standard.

Histogram matching-based semantic segmentation model for crop classification with Sentinel-2 satellite imagery

ABSTRACT Accurate and near-real-time crop mapping from satellite imagery is crucial for agricultural monitoring. However, the seasonal nature of crops makes it challenging to rely on traditional machine learning methods and previous samples generated within specific domains. In this study, we improved the histogram matching method for color correction of multi-temporal images and tested the performance and prediction classification accuracy of three semantic segmentation models based on weak samples. Classification experiments were conducted for nine categories in two cities in Henan province from 2019 to 2022 using 10 m resolution Sentinel-2 images with different feature selection schemes. We trained the models using classified and recorrected results in four selected sites in 2019 and 2020, and designed experiments to assess the performance of the improved histogram matching method and verify the transferability of semantic segmentation models across regions and years. The experimental results showed that the UNet++ model with feature selection and improved histogram matching methods outperformed other models, such as DeepLab V3+ and UNet, in crop classification transfer cases, with better model performance and higher classification accuracy. The UNet++ model without training samples achieved optimal overall accuracy, Kappa coefficient, and mean F1-score values from 2019 to 2022, exceeding 87%, 82%, and 65%, respectively. Moreover, the representative error of weak samples and prediction classification results were analyzed to improve the model robustness. As an application of transfer-learning in crop mapping, the proposed model effectively addressed the classification problem of multispectral satellite imagery with missing labels.

Interpretable machine learning for the identification of estrogen receptor agonists, antagonists, and binders

An abnormal hormonal activity or exposure to endocrine-disrupting chemicals (EDCs) can cause endocrine system malfunction. Among the many interactions EDCs can affect is the disruption of estrogen signalling, which can lead to adverse health effects such as cancer, osteoporosis, neurodegenerative diseases, cardiovascular disease, insulin resistance, and obesity. Knowing which chemical can act as an EDC is a significant advantage and a practical necessity. New Approach Methodologies (NAM) computational models offer a quick and cost-effective solution for preliminary hazard assessment of chemicals without animal testing. Therefore, a machine learning approach was used to investigate the relationships between estrogen receptor (ER) activity and chemical structure to identify chemicals that can interact with ER. For this purpose, the consolidated in vitro assay data from ToxCast/Tox21 projects was used for developing Random Forest classification models for ER binding, agonists, and antagonists. The overall classification prediction accuracy reaches up to 82%, depending on whether the model predicted agonists, antagonists, or compounds that bind to the active site. Given the imbalance in endocrine disruption data, the derived models are good candidates for deprioritising chemicals and reducing animal testing. The interpretation of theoretical molecular descriptors of the models was consistent with the molecular interactions known in the ligand binding pocket. The estimated class probabilities enabled the analysis of the applicability domain of the developed models and the assessment of the predictions’ reliability, followed by the guidelines for interpreting prediction results. The models are openly accessible and useable at QsarDB.org (http://dx.doi.org/10.15152/QDB.259) according to the FAIR (Findable, Accessible, Interoperable, Reusable) principles.

Improvement of data imbalance for digital soil class mapping in Eastern China

A fine-resolution digital soil class map is needed. However, the problem of imbalanced data leads to an inaccurate spatial distribution of the digital soil class map, and the spatial resolution of digital soil class maps at a large scale is low in existing studies. Based on these points, an algorithm of over-sampling and under-sampling was introduced to solve the problem of imbalanced data, and to improve the performance of soil classification model. 316 topsoil samples with eight main soil classes at the great group level were collected in Eastern China. Eight out of twelve prediction variables were determined after the importance evaluation by “Mean Decrease Accuracy” in the random forest (RF) model, including digital elevation model (DEM), enhanced vegetation index (EVI), land surface wetness index (LSWI), land surface temperature (LST), normalized differenced vegetation index (NDVI), and soil texture components. RF model was also applied to complete digital soil class mapping, and the results of treated (over-sampling and under-sampling by randomly increasing or decreasing the number of samples) and untreated data were compared and discussed. Research results indicated that modeling by imbalanced data resulted in uncertain soil classes mapping, with minority classes were lost and with lower accuracies than those of balanced data (overall accuracy = 83.83 %, kappa coefficient = 0.79). After over-sampling and under-sampling treatments, these problems were well solved with an overall accuracy of 96.72 % and a kappa coefficient of 0.93. The accuracy of soil class prediction for minority classes were improved by 12.5 %–54.5 %. Compared to the existing conventional soil map, the new map with a fine resolution of 30 × 30 m is time-effective and more detailed. Validation (point-validation and map-to-map comparison) of the predicted map showed that the output is reliable and can provide a reference for other soil and environmental studies without major difficulties.